- Export on-device chats for import as read-only on other devices. - New models added: Qwen3.6 27B, Qwen3.6 35B A3B, Google's Gemma 4, PrismML's Ternary Bonsai, and Liquid's LFM 2.5 VL. - Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown.

- Export on-device chats for import as read-only on other devices. - New models added: Qwen3.6 27B, Qwen3.6 35B A3B, Google's Gemma 4, PrismML's Ternary Bonsai, and Liquid's LFM 2.5 VL. - Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown.

- Export on-device chats for import as read-only on other devices. - New models added: Qwen3.6 27B, Qwen3.6 35B A3B, Google's Gemma 4, PrismML's Ternary Bonsai, and Liquid's LFM 2.5 VL. - Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown.

- Export on-device chats for import as read-only on other devices. - New models added: Qwen3.6 27B, Qwen3.6 35B A3B, Google's Gemma 4, PrismML's Ternary Bonsai, and Liquid's LFM 2.5 VL. - Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown.

- New models added: Qwen3.6 27B, Qwen3.6 35B A3B, Google's Gemma 4, and PrismML's Ternary Bonsai. - Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown.

- New models added: Google's Gemma 4, Alibaba's Qwen3.6-35B-A3B, and PrismML's Ternary Bonsai. - Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown.

- Gemma 4, Google's most intelligent open models yet, is now live on Russet (E2B/E4B/26B/31B). Gemma 4 outperforms models 20x its size in some industry-standard benchmarks. - Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown.

- Gemma 4, Google's most intelligent open models yet, is now live on Russet (E2B/E4B/26B/31B). Gemma 4 outperforms models 20x its size in some industry-standard benchmarks. - Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown.

- Gemma 4, Google's most intelligent open models yet, is now live on Russet (E2B/E4B/26B/31B). Gemma 4 outperforms models 20x its size in some industry-standard benchmarks. - Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown.

- Gemma 4, Google's most intelligent open models yet, is now live on Russet (E2B/E4B/26B/31B). Gemma 4 outperforms models 20x its size in some industry-standard benchmarks. - Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown.

- Gemma 4, Google's most intelligent open models yet, is now live on Russet (E2B/E4B/26B/31B). Gemma 4 outperforms models 20x its size in some industry-standard benchmarks. - Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown.

- Apple Foundation Model (Apple Intelligence) and several MLX models now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown. - Say hello to the latest and greatest multimodal Qwen3.5 Small and Medium Model Series (SMS) now live on Russet. Download from Settings and spin up a new Chat to try. - Introducing Speech: generate natural and expressive speech from text. As with all else on Russet, neither text nor speech leaves your Mac and processing occurs on-device. - Expanded supported attachment types in chat for all on-device models. Plus new agentic tools thanks to more supported attachment types: Summarize Document and Explain Code.

- Apple Foundation Model (Apple Intelligence) now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. *Tool-calling in Chat requires all permissions (e.g., Calendar access) to be enabled in Agent first so tools can access the data. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown. - Say hello to the latest and greatest multimodal Qwen3.5 Small and Medium Model Series (SMS) now live on Russet. Download from Settings and spin up a new Chat to try. - Introducing Speech: generate natural and expressive speech from text. As with all else on Russet, neither text nor speech leaves your Mac and processing occurs on-device. - Expanded supported attachment types in chat for all on-device models. Plus new agentic tools thanks to more supported attachment types: Summarize Document and Explain Code.

- Apple Foundation Model (Apple Intelligence) now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. *Tool-calling in Chat requires all permissions (e.g., Calendar access) to be enabled in Agent first so tools can access the data. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown. - Say hello to the latest and greatest multimodal Qwen3.5 Small and Medium Model Series (SMS) now live on Russet. Download from Settings and spin up a new Chat to try. - Introducing Speech: generate natural and expressive speech from text. As with all else on Russet, neither text nor speech leaves your Mac and processing occurs on-device. - Expanded supported attachment types in chat for all on-device models. Plus new agentic tools thanks to more supported attachment types: Summarize Document and Explain Code.

- Apple Foundation Model (Apple Intelligence) now supports tool-calling in Chat. You can now simply ask Russet to list today's or tomorrow's schedule directly in Chat instead of going through Agent. *Tool-calling in Chat requires all permissions (e.g., Calendar access) to be enabled in Agent first so tools can access the data. - New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Set your default Persona in Settings > Chat > Persona. Or, better yet, create your own! - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown. - Say hello to the latest and greatest multimodal Qwen3.5 Small and Medium Model Series (SMS) now live on Russet. Download from Settings and spin up a new Chat to try. - Introducing Speech: generate natural and expressive speech from text. As with all else on Russet, neither text nor speech leaves your Mac and processing occurs on-device. - Expanded supported attachment types in chat for all on-device models. Plus new agentic tools thanks to more supported attachment types: Summarize Document and Explain Code.

- New voices in different languages to generate Speech from. Plus a powerful new Speech model: Fish Audio S2 Pro is a leading text-to-speech (TTS) model with fine-grained inline control of prosody and emotion. - Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Available exclusively for MLX models. Set your default Persona in Settings > Chat > Persona. - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown. - Say hello to the latest and greatest multimodal Qwen3.5 Small and Medium Model Series (SMS) now live on Russet. Download from Settings and spin up a new Chat to try. - Introducing Speech: generate natural and expressive speech from text. As with all else on Russet, neither text nor speech leaves your Mac and processing occurs on-device. - Expanded supported attachment types in chat for all on-device models. Plus new agentic tools thanks to more supported attachment types: Summarize Document and Explain Code.

- Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Available exclusively for MLX models. Set your default Persona in Settings > Chat > Persona. - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown. - Say hello to the latest and greatest multimodal Qwen3.5 Small and Medium Model Series (SMS) now live on Russet. Download from Settings and spin up a new Chat to try. - Introducing Speech: generate natural and expressive speech from text. As with all else on Russet, neither text nor speech leaves your Mac and processing occurs on-device. - Expanded supported attachment types in chat for all on-device models. Plus new agentic tools thanks to more supported attachment types: Summarize Document and Explain Code.

- Local Inference Server: Expose Russet's loaded MLX model via an API at localhost, so other apps and tools can use it without loading another model. - Introducing Persona: Tune your response style with 7 presets, including Russet, Study Buddy, and Devil’s Advocate. Available exclusively for MLX models. Set your default Persona in Settings > Chat > Persona. - Benchmark downloaded MLX models in Settings > Developer > Benchmark and save results as Markdown. - Say hello to the latest and greatest multimodal Qwen3.5 Small and Medium Model Series (SMS) now live on Russet. Download from Settings and spin up a new Chat to try. - Introducing Speech: generate natural and expressive speech from text. As with all else on Russet, neither text nor speech leaves your Mac and processing occurs on-device. - Expanded supported attachment types in chat for all on-device models. Plus new agentic tools thanks to more supported attachment types: Summarize Document and Explain Code.

- Say hello to the latest and greatest multimodal Qwen3.5 Small Model Series (SMS) now live on Russet. Download from Settings and spin up a new Chat to try. - Introducing Speech: generate natural and expressive speech from text. As with all else on Russet, neither text nor speech leaves your Mac and processing occurs on-device. - Expanded supported attachment types in chat for all on-device models. Plus new agentic tools thanks to more supported attachment types: Summarize Document and Explain Code.

- Say hello to the latest and greatest multimodal Qwen3.5 Small Model Series (SMS) now live on Russet. Download from Settings and spin up a new Chat to try. - Introducing Speech: generate natural and expressive speech from text. As with all else on Russet, neither text nor speech leaves your Mac and processing occurs on-device. - Expanded supported attachment types in chat for all on-device models. Plus new agentic tools thanks to more supported attachment types: Summarize Document and Explain Code.

- Introducing Voice: generate natural and expressive speech from text in custom voices. As with all else on Russet, neither text nor speech leaves your Mac and processing occurs on-device. - Expanded supported attachment types: .txt, .md, .csv, .tsv, .rtf, .json, .yaml, and source code (.swift, .py, .rs, .js, .ts, .html, .css, .sh, etc.) for all on-device models. Attach a file at the start of a new chat by selecting the paperclip symbol in the input bar. New agentic tools thanks to more supported attachment types: Summarize Document and Explain Code. - New MLX models: Nanbeige4.1-3B (18 GB RAM min.) + Phi 4 mini (16 GB RAM min.) + LFM2.5 1.2B Instruct [(4 & 8 bit) 16 GB RAM min.] + DeepSeek R1 0528 Qwen3 8B [(4 & 8 bit) 18 GB RAM min.] + Gemma 3 12B [(QAT) 24 GB RAM min.] + Ministral 3 3B (16 GB RAM min.) + Qwen3-VL 4B [(6 & 8 bit) 24 GB RAM min.]. - Gemma 3 4B & 12B + Qwen3-VL 4B (6 & 8 bit) models now understand images as input, enabling multimodal chats. - New Space theme in Settings for pitch black aesthetic. - Performance optimizations across the board for increased speed and reliability. Improved inline markdown parsing and LaTeX rendering.

- Redesigned background for style, MLX models manager for simplicity, Agent results container for sleekness, and Settings window for smoothness. Plus, 3 new fonts for you to choose from in Settings: Lato, Noto Sans, and Ubuntu! - New MLX models: LFM2.5 1.2B Instruct [(4 & 8 bit) 16 GB RAM min.] + DeepSeek R1 0528 Qwen3 8B [(4 & 8 bit) 18 GB RAM min.] + Gemma 3 12B [(QAT) 24 GB RAM min.] + Ministral 3 3B (16 GB RAM min.) + Qwen3-VL 4B [(6 & 8 bit) 24 GB RAM min.]. - Added speed and intelligence ratings for MLX models to help you pick the best model for your chats. The ratings are just based on my opinions so use them as a friendly guide. - Gemma 3 4B & 12B + Qwen3-VL 4B (6 & 8 bit) models now understand images as input, enabling multimodal chats. - View Russet's resource usage (RAM, CPU, GPU) in real-time using the new Sidebar Metrics panel. - Performance optimizations across the board for increased speed and reliability. Improved inline markdown parsing. And no more approximations for token counts in chats with MLX models, as the exact number of tokens consumed is now shown.

- Redesigned Agent results container for sleekness and Settings window for smoother navigation and enhanced model discovery. - New MLX models: Gemma 3 12B [(QAT) 24 GB RAM min.] + Ministral 3 3B (16 GB RAM min.) + Qwen3-VL 4B [(6 bit) 24 GB RAM min.] + LFM2.5 1.2B Instruct [(4 & 8 bit) 16 GB RAM min.] - Added speed and intelligence ratings for MLX models to help you pick the best model for your chats. The ratings are just based on my opinions so use them as a friendly guide. - Gemma 3 4B & 12B + Qwen3-VL 4B models now understand images as input, enabling multimodal chats. - View Russet's resource usage (RAM, CPU, GPU) in real-time using the new Sidebar Metrics panel. - Performance improvements and optimizations across the board for increased speed and reliability.

- Added speed and intelligence ratings for MLX models to help you pick the best model for your chats. The ratings are just based on my opinions so use them as a friendly guide. - View Russet's resource usage (RAM, CPU, GPU) in real-time using the new Sidebar Metrics panel. - Shorten URLs and find words related to your input using the new Shorten URL and Find Related Words Web tools in Agent. - Add events to your system from Russet using the new Create Event tool in Agent. - Translate text quickly and without needing the Internet using the new Translate function in Flow. Powered by Apple Intelligence, input language is auto-detected and translated to your choice of target language. Also added text-to-speech support for Flow outputs. - Improved table and inline code rendering in Chat. Also quashed some bugs, improved overall UI & stability, and boosted performance.

- View Russet's resource usage (RAM, CPU, GPU) in real-time using the new Sidebar Metrics panel. - Shorten URLs and find words related to your input using the new Shorten URL and Find Related Words Web tools in Agent. - Add events to your system from Russet using the new Create Event tool in Agent. - Translate text quickly and without needing the Internet using the new Translate function in Flow. Powered by Apple Intelligence, input language is auto-detected and translated to your choice of target language. Also added text-to-speech support for Flow outputs. - Improved table and inline code rendering in Chat. Also quashed some bugs and improved overall stability.